Portal de Periódicos da CAPES

Bases of Jerzy Konorski's theory of synaptic plasticity

2019; Wiley; Volume: 51; Issue: 9 Linguagem: Inglês

10.1111/ejn.14532

1460-9568

Łukasz Bijoch, Małgorzata Borczyk, Rafał Czajkowski,

Neurological Disorders and Treatments

The first half of the 20th century was a time of creation, conflict and unification attempts of major theories in neuroscience (Zieliński, 2006). There were physiologists, inspired by Charles Scott Sherrington (1857–1952), who claimed that studies of nerves should be at the forefront of scientific research and were key to understanding the brain (Sherrington, 1906). They alleged that cognitive processes could be represented as adaptations in neural pathways. Alternatively, behaviorists such as Burrhus Frederic Skinner (1904–1990), who, although inspired by Sherrington's work, stressed that behavior should be the focus. They argued that behavior should be represented as an input (stimulus) and output (action) mechanism and constitute a separate area of study (Catania & Laties, 1999; Skinner, 1966). At the time, many Russian neuroscientists focused around Ivan Pavlov (1849–1936), who was studying classical conditioning (also known as Pavlovian conditioning), defined as an experimental paradigm whereby an unconditioned response (such as salivary response to food presentation) is paired with a previously unrelated stimulus (such as a ring of a bell) (Konorski, 1948, 2013; Pavlov, 1902). His groundbreaking research was bolstered by scientists of the Pavlovian school who performed many detailed experiments in this field and as Konorski mentioned in his book, their research was a precious collection. However, it was published mostly in Russian. Due to the language barrier, it was available only to a small group of contemporary western scientists (Konorski, 1948). The physical and linguistic impediments of neuroscientific groups was a pivotal hindrance in the fundamental understanding of how the brain performs cognitive functions. Under such circumstances, two independent scientists had a crucial impact on the development of neuroscience. First was Jerzy Konorski (1903–1973), who was known for his discovery of conditioned reflexes type II (later called also "instrumental reflexes") and second Donald Hebb, an acclaimed expert in the field of human and animal learning (1904–1985) (Berlucchi & Buchtel, 2009; Brown, 2007; Wyrwicka, 1994). They both demonstrated a link between higher mental capabilities and neural activity. As a result, Konorski in 1948 and, almost parallelly, Hebb in 1949 described a theory of cellular mechanism of learning – synaptic plasticity, defined as a change in connection quality between neurons (Hebb, 1949; Kandel, 2012; Konorski, 1948). They related the strengthening of synapses with creation engrams (encoded memories) and their weakening with forgetfulness. Their theories share many similarities, however, Konorski particularly emphasized the idea of pre-existing neuronal connections between distinct brain regions. He claimed that neural plasticity is not simply the establishment of new connections between neurons, but a remodeling of presently existing synapses, created during ontogenesis. This revolutionary thought and theorized principles put forward by Konorski was based on observations made during his own (and Pavlovian school) classical conditioning experiments. Here, learning appeared to occur rapidly and conditioned responses (CRs) could form after as little as two trials. It was inconceivable that newborn connections arose with that speed between spatially distant centers of the brain (Konorski, 1948). Today, the concept of structural and functional neuronal plasticity as fundamental processes in learning is widely accepted, but in the first half of the 20th century, it was not. Even though such hypotheses were formulated at the end of 19th century (by Wiliam James and Eugenio Tanzi), as Konorski wrote in his book, the idea of synaptic plasticity was seldom studied (Berlucchi & Buchtel, 2009; Konorski, 1948). The Pavlovian school was investigating behavioral details of conditioning and showed little interest in the physiological consequences of it (Konorski, 2013). A lack of compromise between physiologists and behaviorists existed, resulting in little attention to the molecular principles of learning (Konorski, 1948). In retrospect, Jerzy Konorski stating that the observed CR (and learning in general) was a function of synaptic processes was truly radical for the times. The association between the synaptic and behavioral adaptations helped to explain many questions regarding brain function and started a novel branch of science: cellular neurobiology (Nicholls, In Martin, Wallace, & Kuffler, 1992). Many decades after Konorski had formulated his hypothesis, experimental neuroscience began to provide evidence that synaptic plasticity was indeed the process underlying learning (Kandel, 1981). A sophisticated balance between strengthening and weakening of synaptic connections does appear to exist. One of the most striking observations in the field of cellular-based learning was made by Eric Kandel in his now-famous experiments of defensive reflexes in the snail Aplysia californica. However, these experiments were made decades after Konorski had formulated his ideas (Kandel, 1965; 1978; 1979). Jerzy Konorski's input into cellular neuroscience was not only theoretical, as he claimed to have experimental evidence supporting his theory about the presence of pre-existing connections between neurons (Dobrzecka & Konorski, 1962; Dobrzecka et al., 1965). In later works, his pupils referenced those experiments while confirming Konorski's early ideas (Zieliński, 2006). Unfortunately, due to the collapse of Polish science after War World II, the idea of pre-existing connections described in 1948 had to wait almost twenty years to be experimentally confirmed (Wyrwicka, 1994; Zieliński, 1994). It is worth mentioning that after World War II not only buildings and scientific resources were in deplorable conditions. Poland was also struggling because of a lack of scientists (who died or emigrated to other countries) (Wyrwicka, 1994; Zieliński, 1994). Proof of the pre-existing neuronal connections concept was presented in two above-mentioned articles from 1962 to 1965, together with Czesława Dobrzecka and Barbara Sychowa (Dobrzecka & Konorski, 1962; Dobrzecka et al., 1965). Despite the high number of citations of Konorski's book (over 1,000), these papers remain poorly acknowledged and have reached respectively 11 and 32 citations, mostly by Konorski's students and colleagues. One possible explanation is the limited transfer of information to western societies during the cold war. Other works of Konorski from that period, published with western collaborators on the separation between instrumental and salivary responses in conditioning are more recognized (Ellison & Konorski, 1964, 1965). To this day, these articles remain poorly known, perhaps due to inaccessibility and/or their lack of clarity. Thus, in this article, we refresh these articles by reviewing their pertinent findings and display the data in modern form. Our goal is to make Konorski's work more accessible and to show the significance and impact of his theory of plasticity on the development of modern neuroplasticity studies. In the aforementioned experiments, the authors used a model of operant conditioning (also known as instrumental or type II conditioning), which is a learning process through which the behavior is modified by reinforcement or punishment (Miller & Konorski, 1928; Skinner, 1938). During the experimental procedure, a stimulus is presented during the reward or punishment period and, with time, the stimulus alone can trigger such behaviors. The article, "On the peculiar properties of the instrumental conditioned reflexes to 'specific tactile stimuli'" from 1962, was one of the first to describe that distinct modalities can create similar operant responses in dogs. In the introduction, researchers stated their hypothesis: "The supposition was put forward that motogenic properties of stimuli might also depend on their intrinsic character; in particular, it was suggested that the tactile distal part of the leg involved in the performance of the trained movement might appear to be more motogenic than other CSi. Furthermore, learning is considerably faster and the CR is harder to extinguish if the stimulus is applied to the same limb used for performing the instrumental response. This may be explained by the presence of pre-existing connections between brain regions representing limb, which should be especially abundant, if located in a close proximity (Dobrzecka & Konorski, 1962). In the second article, "The effects of lesions within the sensorimotor cortex upon instrumental response to the 'specific tactile stimulus'" from 1965, the authors describe the effects following ablating direct connections between closely related regions involved in operant conditioning, namely the sensory and motor areas of the cortex. Such lesions impair the response, but they do not inhibit them completely. Authors claimed that after the surgery instrumental response may still be possible due to an indirect pathway connecting these regions of the cortex. This pathway is presumably based on pre-existing connections (Dobrzecka et al., 1965). These articles complement each other, as the first one characterizes the behavior, whereas the latter tries to confirm the hypothesis by manipulating connections in the brain via lesioning. In the following two segments of this review, we discuss experimental methods and results presented by Konorski and his colleagues from these two selected publications. The article "On the peculiar properties of the instrumental conditioned reflexes to 'specific tactile stimuli'" from 1962 is the first report supporting Konorski's theory that learning is based on changes of pre-existing connections (Konorski, 1948). A scheme of operant conditioning, called by Konorski II-type conditioning, was applied as a model of learning (Figure 1b). Conditioning was performed on stray dogs of undocumented sex and weight. Upon presentation of the conditioning stimulus, animals were trained to perform an action. Specifically, the task was to place the right paw on a tray to obtain a food reward. The whole system was organized in a soundproof chamber so that the dog was separated from the experimenter. Upon reward presentation, an unconditioned response (salivation) was measured in selected animals. Depending on the experiment, some of the dogs were trained to a specific tactile, an auditory (buzzer or metronome) or a non-specific tactile (to the side of the body) stimulus. Specific Tactile Stimulus (STS) was called specific as it was applied to the same paw the animal had to raise to obtain a reward, in contrast to the non-specific tactile stimulus applied to a distal body part. Konorski's hypothesis was that STS would be easier to train and harder to extinguish (forget) due to the physical proximity of the sensory and motor regions representing the same limb. This hypothesis was based on the idea that, before the training started, connections between sensory and motor areas existed and those representing the same leg were already strong. training also transfer, extinction and recall paradigms were performed to distinguish different properties of the tested stimuli. In the first experiment described in the article, animals were trained to the STS, pinning of the right foreleg (Figure 1b). During the initial training phases, experimenters passively placed the dog's paw on the tray. Once the conditioning responses were well established, which required less than 20 trials in all cases, researchers attempted to transfer the CR to a new stimulus – an auditory tone (Figure 2a). This auditory signal appeared 2–3 times in every experimental session among the tactile stimulus (Figure 2a.i–a.iii), before testing, to determine if they initiated the action upon auditory stimulus (buzzer) alone (Figure 2b.i–b.ii). This transfer was very inefficient, despite the high number of transmission trials (Table 1, dogs a–f). Only one out of five dogs established CRs to a new stimulus. However, despite the lack of instrumental conditioning, dogs did learn something during transmission experiments. Classical conditioning arose during transfer trials and was manifested by a salivation reflex after the buzzer noise. Thus, an association between the buzzer and the reward did indeed occur. The "peculiar nature" of the STS arose from a series of control experiments. Here, the transfer of the instrumental response was easily possible if the first stimulus was not STS, but the auditory stimulus (Table 1, dogs g–i) or abdominal pinning stimulation (Table 1, dogs j & k). These results demonstrate that when a stimulus is administered to the same limb that is used to generate a response, a specific, more rigid conditioning occurs, which is different than in all other cases. This, according to Konorski, indicates a different nature of pre-existing connections between the sensory and motor areas of the same limb than between auditory, or sensory areas of another part of the body and motor area of the right limb. Next, the authors tested for differences in the stability and persistence of operant conditioning to this peculiar STS. To this end, they performed extinction and restoration sessions to the auditory stimulus and to the STS (Figure 3). Moreover, as a positive control, another auditory stimulus – metronome was introduced. Thus, all four animals of the experiment were trained to three stimuli of which two (STS and buzzer) were subjected to chronic extinction and restoration protocols, while the metronome was always reinforced with rewards. Involuntary responses (salivation) were measured throughout the sessions. In the extinction series, the metronome was applied with reinforcement seven times per session, while buzzer and STS were applied only once per session, without reinforcement, in the 3rd (or 4th) and 6th (or 7th) trial. The order of buzzer and STS alternated between sessions. Extinction meant that the dogs were no longer raising their right paw onto the tray neither after the buzzer sound nor after the STS. Following extinction, trials with reinforcement were introduced again. Quantification of the number of trials needed to extinguish, thus to remodel the memory, was greater for STS (Table 2). Extinction to buzzer appeared faster than to STS. Instrumental responses to the auditory stimuli rarely restored (only one success), as compared to STS (restored relatively fast after two trials in 100% dogs). The above experiments show that the quality of association in instrumental conditioning depends also on the type of stimulus and not only, as Pavlov claimed, on the strength. To be precise, the Pavlovian "stronger the stimulus, stronger the response" relation holds true for involuntary, salivary reactions and is discussed in the research paper. The authors state: "Thus, the old and well documented Pavlovian principle of the dependence of CR strength on the CS strength should be reformulated by stating that for different types of CRs there exists a different hat hierarchy of strength" (Dobrzecka & Konorski, 1962). The main conclusion from the results presented above is that instrumental conditioning is stronger and easier to restore for STS than for an auditory tone. This, according to Konorski, results from the diverse strengths of pre-existing connections "Now, one may suppose that the exceedingly strong motogenic of the specific tactile stimulus is due to the powerful direct connections existing between the centre of this stimulus and the relevant moter centre." Three years after providing evidence that auditory and STS conditioning have different properties, Konorski and Dobrzecka published a research article entitled: "The effects of lesions within the sensorimotor cortex upon instrumental response to the 'specific tactile stimulus'" where they describe the effects on the CR, following lesioning regions involved in this type of learning. The authors state the rationale behind these experiments to be the following: "It has been shown that all these peculiar properties of the STS may be explained by an assumption that the connections between the STS center and the center of the corresponding motor act are stronger than the connections between other CSi and the latter center. This increased strength of the connections was attributed to the organization of the sensorimotor cortex in which the sensory area representing the given limb lies in close vicinity to the corresponding motor area and these two areas are presumably interconnected" (Dobrzecka et al., 1965). Therefore, they undertook an approach using cortical lesions. Mongrel dogs (1.5–3 years old, 10–18 kg) were used in these experiments. Based on the organization of the cortex, the authors chose to make incisions between the somatosensory and motor cortices or lesioned the sensory motor or motor area. As the latter of these two manipulations had a limited number of animals, and protocols for both the surgery and behavioral tasks were not consistent, we chose to discuss only the incision experiments in greater detail. Parts of the cortex were distinguished based on sulci and gyri. Somatosensory and motor cortices, representing the foreleg, lie very close in proximity and can be easily separated by an incision made in the sulcus centralis (Figure 4). All surgeries were performed contralaterally to the leg used in the experiment. Four dogs were trained before the surgery and underwent extinction and restoration protocols, as described in the experiments above. For two dogs (Table 3, dogs b & d), extinction and retrieval procedures were both performed before and after the surgery, while in the other two dogs, these procedures occurred only after the surgery (Table 3, dogs a & c). The goal of these trials was to impair the resistance of STS to extinction. Similar to the previous experiment, measurements of extinction were the number of trials, which were the first of three consecutive, no-response trials. Surgeries had a limited effect on general animal performance, as dogs were capable of freely moving without visible difficulties. Moreover, lesions had little effect on already existing instrumental CRs. However, the authors observed that the incision affected buzzer conditioning, which became more stable than before the surgery (Table 3), while extinction to the STS became easier (Table 3). These changes were interpreted to result from the severing of inhibitory connections, that were strengthened during initial training (before the lesion). The consequences of removing the motor or sensory cortices on the extinction of the instrumental response of STS were similar (Dobrzecka et al., 1965, Table II and III) to those obtained with an incision between these two regions. Thus, in their discussions that have stated: "In our previous paper it was supposed that the peculiar properties of the instrumental CR to the STS are due to the close intercortical relations existing between the sensory and the motor area of the cerebral cortex. This hypothesis brought us to the idea that the separation of these two areas by simple incision penetrating into the white matter might deprive the STS of its peculiar properties and convert this stimulus into a regular CS of an instrumental reflex. This hypothesis has been fully confirmed by decrescent experiments". In experiments published in 1962, Konorski and Dobrzecka demonstrated that STS elicits CRs that are more stable than those associated with other, non-specific, stimuli. Data from 1965 indicates that specific properties of STS on CRs are disrupted if close-related connections between brain regions putatively operating STS and CR are destroyed. Moreover, experiments with lesions showed that it is possible to convert specific instrumental CRs to a standard one, namely to make the STS as easy to extinct as an auditory stimulus. These results suggest that pre-existing indirect connections participate in the formation of the CR (but without lesion, they are masked by a stronger pathway). Due to an indirect pathway, the conditioning dependent on these connections becomes less stable. In the discussion section, the authors summarize: "The incision between the two areas […] changes the particular structure of the instrumental CR arc to the STS by depriving it of one branch responsible for its particular properties" (Dobrzecka et al., 1965). According to Konorski's hypothesis, if such pathways did not exist, CRs should disappear completely after lesioning the direct route (Figure 5). With these findings, Konorski again stood away from the Pavlov school, which had claimed that the strength of the conditioning is based solely on the strength of the stimulus (Konorski, 1948). This was a pattern in his works, as many of his experimental results suggested that some of Pavlov's assumptions were incorrect. It caused Konorski many quandaries, as he was completely ostracized by the mainstream scientists from the Soviet Union for years, which in turn had cost him academic positions (Wyrwicka, 1994). To read more about his life, we recommend memorials written by Konorski's collegues and his autobiography (Konorski & Lindzey, 1974; Wyrwicka, 1994; Zieliński, 1994, 2006). Here, the authors have shown that the strength of conditioning also depends on the type of conditioning and the nature of stimuli used. Characteristics of pre-existing connections thus define the valence and strength of signals from each of the cortical areas involved in stimulus processing and animal performance. It is believed that these connections arise during ontogenesis, but they become functional only during the experimental manipulation. The strong effect of the STS was thought to be due to the proximity of regions processing the sensory input and movement of the same limb. However, later experiments from Konorski's group shown that if the dog has food ad libitum, it blocks the conditioning process, which indicates there may be some affective context for this form of learning to occur, presumably provided by food deprivation in this case (Dobrzecka & Wyrwicka, 1960). Indeed, it is now known that circuity participating in such conditioning is more complicated and involves subcortical structures, like the amygdala, a structure associated with emotional processing (Knapska et al., 2013). Moreover, it is now established that many other cortical and limbic brain regions play critical roles in associative learning (Bocchio, Nabavi, & Capogna, 2017; Tonegawa, Morrissey, & Kitamura, 2018). Konorski, Dobrzecka and Sychowa had very limited tools for their studies. The variety and low numbers of animals used in their experiments raise concerns about the quality of this proof of concept. Despite the fact that their findings may be explained by the Konorski theory of pre-existing connections, these experiments only provide preliminary evidence that such associations are evident. With the advent of new and sophisticated techniques, confirmation of his early theory of pre-existing connections playing a crucial role in learning has transpired. This idea of pre-existing connections created during ontogenesis has survived the trial of time. It is now accepted that such connections arise during development and become stronger or weaker, depending on the experience (Hyson & Rudy, 1984; Stiles & Jernigan, 2010). During this time, Konorski did not know if every center of the brain was interconnected. Experimental evidence from this era indicated that in theory, any stimuli could elucidate conditioning, however, sometimes with obstacles (Konorski, 1948). Presently, we cannot exclude this idea. In fact, some pre-existing connections may remain inactive throughout life. The phenomenon of phantom limb, occurring after limb amputation or other phantom senses may be based on these connections and their ability for plasticity and reorganization (Flor, Nikolajsen, & Jensen, 2006). Konorski's innovative ideas and fundamental experiments supporting his theory of pre-existing connections initiated new research topics and provided a better understanding of synaptic plasticity. The concept of plastic changes is specifically organized within the brain network, where memory is created, prompted the development of new tools to study learning. With markers of neuronal plasticity such as expression of immediate early genes (c-fos and alike), functional changes in ion currents (measured with patch-clamp electrophysiology) or ultrastructural changes in neuronal morphology (such as growth or pruning of dendritic spines), it became possible to physically measure and observe elementary parts of memory engrams (Czajkowski et al., 2014; Nikolaev, Kaczmarek, Wei, Winblad, & Mohammed, 2002; Shukla et al., 2017; Vetere et al., 2011). Importantly, novel techniques allow for distinction and selective manipulation of neuronal subsets (such as excitatory or inhibitory neurons), which provide even more insight into the synaptic bases of memory formation. Molecular pathways of synaptic plasticity have been elucidated in considerable detail and it is now possible to modulate engrams in animals by direct alterations of neuronal signaling (Etkin et al., 2006; Kauer & Malenka, 2007; Knapska et al., 2013; Tang et al., 1999). The concept that projections in the brain are created during ontogenesis set forth the rapid development of another branch of neuroscience: connectomics. Konorski, in his monograph from 1948, distinguished two types of neurons creating potential connections: those which send signals out and those which receive signals in. This dogma is central to connectomics and predicts signal processing by the brain. Currently, we can track these connections by retro- and anterograde viral vectors in a number of animal species (Ugolini, 2010). As a result, we now have at our disposal, a number of databases of human, mouse and Drosophila brain atlases, such as those developed by the Allen Brain Institute (alleninstitute.org; virtualflybrain.org), to provide us with unprecedented detail about these connections. Recently, a connectomic approach was combined with optogenetics, whereby a viral vector carrying information about the light-sensitive ion channel was injected into one structure, but only expressed in cells, which project to another part of the brain, where a second viral vector has been introduced (Häusser, 2014; Zemelman, Lee, Ng, & Miesenböck, 2002). The existence of these connections, firstly, makes it possible to equip them with light-sensitive channels and secondly, to activate them with light in order to manipulate a specific memory. Such methods may be used to block or enhance learning, affect recall and even create "false memories" in an animal's brain (Assareh, Bagley, Carrive, & McNally, 2017; Nabavi et al., 2015; Ramirez et al., 2013). By studying Jerzy Konorski's work and ideas, we come to realize that the wisdom he bestowed to the theory of synaptic plasticity, especially the aspect of pre-existing connections, may be deemed to be his most important contribution to neuroscience. What were his own thoughts about his influence? Some light may be shed by a text found in 2013 in the library of The Nencki Institute. This article, entitled: "Study of behavior: Science or pseudoscience" dated around 1970 had not been published until 2013. In this critical review, he shares his personal view on different approaches to studying neuroscience. In this article, around fifty years ago, Konorski wrote: "Although some decades ago psychologists could claim that the skull was a black box and that we could not even imagine what is going on inside, now this black box is illuminated by the light coming from direct penetration into it either by the surgical knife, or drugs, or implanted electrodes. Therefore, our long-lasting dream of fixing correspondence between mental processes and cerebral processes does not seem unrealistic and is even not very far from realization." We must admit that we still do not understand the mind truly; however, progress has been made in understanding how the brain functions. Partially thanks to Konorski, who thought that by finding a comprehensive approach to studying the brain, one may establish a true connection between different branches of neuroscience, and global progress would be possible. From his 1948 book, the manuscript from around 1970 and finally from his autobiography, we can assume that he felt he reached this goal satisfactorily by the end of his scientific career (Konorski, 1948; 1974; 2013). His last big contribution to neuroscience was his second monography, published in 1967. In the book, he summarized how neuronal activity was understood at the time and formulated new theories about brain activity. One of his most striking theories was the concept of gnostic cells, which claimed that a representation of an object or abstraction is specifically encoded by group of neurons in the brain (Konorski, 1967). Similarly to the description of instrumental conditioning and the theory of the synaptic plasticity, hypothesis of gnostic cells was also independently formulated by another western scientist: Jerry Letvin (Gross, 2002; Quiroga, 2013). Nowadays, such neurons encoding specific concept are also known as "grandmother cells" and, at least partially, this phenomenon seems to correctly predict how brain responds to complex stimuli from external world. One of the most striking and famous evidences for it was the discovery of "Jennifer Anniston cell": a single neuron found in human brain, which responded specifically to a picture strictly related to this and no other actress (Quiroga, 2005, 2013). Konorski is often presented as the man responsible for coining the term "synaptic plasticity": the ability of neural networks to rearrange, in order to create a behavioral response (Lamprecht & LeDoux, 2004). This term, idea, and phenomenon appear to provide a remarkable link between behavior and the physiological aspects of one of the higher mental capabilities: learning. However, surprising as it may be, the theory had to wait almost 20 years to be experimentally confirmed. Even after the release of the discussed publications, we can assume that this proof of concept was not well known (evidenced by the low citation numbers in the literature). It is also hard to find attempts to affirm the idea of pre-existing connections by other researchers from the same era. However, many of their experiments were based on this presumption. Today we have strong evidence from multiple studies that supports the idea of plastic, pre-existing connections. The use of optogenetics, electrophysiological recordings, and other commonly used techniques are possible because the actual synapses are already present. We hope that this review has given the reader a journey back in time, to get a glimpse into the first proofs of the theory of pre-existing connections undergoing plasticity. MB and LB contributed equally to the manuscript. Authors thank Dr. Sheldon Michaelson for his valuable proofreading and Prof. Andrzej Wróbel and Prof. Leszek Kuźnicki for their important comments on this project.